Spring Batch 从同一执行和步骤重新启动未完成的作业

Question

我使用以下逻辑重新启动未完成的 Spring Batch（例如应用程序异常终止后）作业：

public void restartUncompletedJobs() {

    LOGGER.info("Restarting uncompleted jobs");

    try {
        jobRegistry.register(new ReferenceJobFactory(documetPipelineJob));

        List<String> jobs = jobExplorer.getJobNames();
        for (String job : jobs) {
            Set<JobExecution> runningJobs = jobExplorer.findRunningJobExecutions(job);

            for (JobExecution runningJob : runningJobs) {
                runningJob.setStatus(BatchStatus.FAILED);
                runningJob.setEndTime(new Date());
                jobRepository.update(runningJob);
                jobOperator.restart(runningJob.getId());
                LOGGER.info("Job restarted: " + runningJob);
            }
        }
    } catch (Exception e) {
        LOGGER.error(e.getMessage(), e);
    }
}

这工作正常，但有一个副作用 - 它不会重新启动失败的作业执行，而是创建一个新的执行实例。如何更改此逻辑以便从失败的步骤重新启动失败的执行并且不创建新的执行？

更新

当我尝试以下代码时：

public void restartUncompletedJobs() {
try {
    jobRegistry.register(new ReferenceJobFactory(documetPipelineJob));

    List<String> jobs = jobExplorer.getJobNames();
    for (String job : jobs) {

    Set<JobExecution> jobExecutions = jobExplorer.findRunningJobExecutions(job);

    for (JobExecution jobExecution : jobExecutions) {
        jobOperator.restart(jobExecution.getId());
    }
    }
} catch (Exception e) {
    LOGGER.error(e.getMessage(), e);
}
}

它失败，但有以下异常：

2018-07-30 06:50:47.090 ERROR 1588 --- [           main] c.v.p.d.service.batch.BatchServiceImpl   : Illegal state (only happens on a race condition): job execution already running with name=documetPipelineJob and parameters={ID=826407fa-d3bc-481a-8acb-b9643b849035, inputDir=/home/public/images, STORAGE_TYPE=LOCAL}

org.springframework.batch.core.UnexpectedJobExecutionException: Illegal state (only happens on a race condition): job execution already running with name=documetPipelineJob and parameters={ID=826407fa-d3bc-481a-8acb-b9643b849035, inputDir=/home/public/images, STORAGE_TYPE=LOCAL}
    at org.springframework.batch.core.launch.support.SimpleJobOperator.restart(SimpleJobOperator.java:283) ~[spring-batch-core-4.0.1.RELEASE.jar!/:4.0.1.RELEASE]
    at org.springframework.batch.core.launch.support.SimpleJobOperator$$FastClassBySpringCGLIB$$44ee6049.invoke(<generated>) ~[spring-batch-core-4.0.1.RELEASE.jar!/:4.0.1.RELEASE]
    at org.springframework.cglib.proxy.MethodProxy.invoke(MethodProxy.java:204) [spring-core-5.0.6.RELEASE.jar!/:5.0.6.RELEASE]
    at org.springframework.aop.framework.CglibAopProxy$DynamicAdvisedInterceptor.intercept(CglibAopProxy.java:684) [spring-aop-5.0.6.RELEASE.jar!/:5.0.6.RELEASE]
    at org.springframework.batch.core.launch.support.SimpleJobOperator$$EnhancerBySpringCGLIB$$7659d4c.restart(<generated>) ~[spring-batch-core-4.0.1.RELEASE.jar!/:4.0.1.RELEASE]
    at com.example.pipeline.domain.service.batch.BatchServiceImpl.restartUncompletedJobs(BatchServiceImpl.java:143) ~[domain-0.0.1.jar!/:0.0.1]

以下代码在作业存储数据库中创建新的执行：

public void restartUncompletedJobs() {
try {
    jobRegistry.register(new ReferenceJobFactory(documetPipelineJob));

    List<String> jobs = jobExplorer.getJobNames();
    for (String job : jobs) {

    Set<JobExecution> jobExecutions = jobExplorer.findRunningJobExecutions(job);

    for (JobExecution jobExecution : jobExecutions) {

        jobExecution.setStatus(BatchStatus.STOPPED);
        jobExecution.setEndTime(new Date());
        jobRepository.update(jobExecution);

        Long jobExecutionId = jobExecution.getId();
        jobOperator.restart(jobExecutionId);
    }
    }
} catch (Exception e) {
    LOGGER.error(e.getMessage(), e);
}

}

问题是 - 如何在应用程序重新启动后继续运行旧的未完成的执行而不创建新的执行？

Answer 1

TL;DR：Spring Batch 将始终创建新的作业执行，并且不会重用之前失败的作业执行来继续执行。

更长的答案：首先你需要了解 Spring Batch 中三个相似但不同的概念：Job、Job Instance、Job Execution

我总是用这个例子：

• 作业：日终批处理
• 作业实例：2018-01-01 的日终批处理
• 作业执行：2018 年 1 月 1 日的日终批处理，执行 #1

概括地说，这就是 Spring Batch 的恢复工作原理：

假设您的第一次执行在步骤 3 中失败。您可以使用相同的参数 (2018-01-01) 提交相同的作业 (End-of-Day Batch)。Spring Batch 将尝试查找提交的作业实例(End-of-Day Batch for 2018-01-01) 的最后一个Job Execution (End-Of-Day Batch for 2018-01-01, execution #1 )，并找到之前在第 3 步中失败。 Spring Batch 然后将创建一个新的执行，[End-Of-Day Batch for 2018-01-01, execution #2]，并从第 3 步开始执行。

因此，按照设计，Spring 试图恢复的是先前失败的作业实例（而不是作业执行）。当您重新运行之前失败的执行时，Spring 批处理不会重用执行。